Picture this: You’re in the middle of a board meeting, tensions are already high, and suddenly, one of your board members raises a question: “Did we confirm the data on that last slide?” 

That particular slide was produced by the product team, who are usually reliable. The data looks solid but you aren’t sure if it was verified. You’re not even certain where they got the data from. The board member shows you their phone with different, contrasting data from a trustworthy source. 

 Turns out, the team got their information from ChatGPT and neglected to verify it. Now, your credibility as a founder is at stake. People are looking at you as if you’re the next Elizabeth Holmes and you’re trying to explain that this one slide got it wrong, but how can you even be sure? Is the entire deck produced by AI? Did we revise any of this information? We have financial data, marketing, and stats – everything is in here. 

AI presents challenges to companies. And it’s not just about misinformation being sourced from the chatbots; it’s also about using real information in the wrong kind of way. 

Teams could also be using accurate data the wrong way

Prosecraft was founded on the premise that all information on the web is fair use, even information found on pirated websites. 

Its founder, Benji Smith, didn’t buy the books that his technology analyzed. Instead, he used information from websites that pirated books. Even though his tech was useful for writers and could have made a positive impact in the publishing world, he ended up closing business, taking down a library of 27,000 titles. He didn’t intend for his AI startup to infringe on copyright or cause such a revolt among what would be, if things had gone differently, his raving customers. 

The problem was that he didn’t buy the books nor did he get permission from authors to use data pulled from their work – and all this at a moment when the world is demanding regulations on AI data. Such demands are coming especially from global creatives,including members of the Writers Guild of America, who have recently ended a 5-month strike based on AI regulations. 

Some data shared by chatbots may be subject to fair use under copyright law –  but not all of it. Some authors and creatives are against their work being used to train the language models. For ChatGPT users, this means that just because you found the information online, doesn’t  mean that it belongs to you.

For ChatGPT users, this means that just because you found the information online, doesn’t mean that it belongs to you. 

While not every product or company runs this same risk, there are still perils in using chatbots without regulations. Think of the Samsung team who uploaded confidential code to ChatGPT, making it available to the internet. That incident resulted in a company-wide ban of ChatGPT.

This is why having company policies on the use of ChatGPT has become so crucial. What are the expectations and how should team members make decisions that involve data gathering from chatbots? 

Companies need to self-regulate their chatbot policies

Government regulations are currently in the developmental stage. As time progresses, there is a growing demand for more regulations from figures such as prime ministers, CEOs, and executives. For instance, the editor in chief of Wired Magazine has called for “AI overlords” to collaborate with governments in regulating data and AI interconnections due to concerns about potential threats to democracy. 

Despite these calls for regulation, the responsibility for determining how to best utilize tools like ChatGPT within their specific industry ultimately lies with each individual company.

Benefits of using chatbots in startups

Even though there are countless cautionary tales revolving around the use of chatbots, a company shouldn’t disregard the use of these tools.

ChatGPT can be useful when researching topics, getting to know your audience better, drafting letters, brainstorming new ideas, or even rewriting code. Some companies use chatbots for customer support, doing market research, and even language translation. OpenAI has Dall E (​​it can “create realistic images and art from a description in natural language”) and Whisper (an “automatic speech recognition system”), so it’s not just text-based outputs that are available to you and your team –  image and voice outputs can be accessed as well. 

Leaders should be open with their teams, engaging them in conversation about the appropriate use of large language models as these are adopted and applied more widely in companies. In this article, we’ll discuss the importance of talking about these topics with your team sooner rather than later. 

Preparing for a Discussion on the Use of ChatGPT: Key Considerations

Adequate preparation is crucial when it comes to talking with  your team about the appropriate and skillful use of ChatGPT. Here are essential considerations to keep in mind before discussing ChatGPT’s use with your team:

1. Define your leadership expectations: The first step is to have clarity on what your expectations are. What are your specific goals and objectives with respect to AI use? Are you aiming to enhance productivity, mitigate risks, or foster innovation? Equally important is identifying the boundaries and limitations on ChatGPT’s usage.

Determine how you envision AI tools fitting into your team’s workflow and decision-making processes.

Adopting the 5W approach (Who, What, When, Where, Why, How) can guide you in crafting comprehensive ChatGPT guidelines that align with your leadership goals.

2. Assess your team’s needs and interests: Understanding your team’s specific interests and requirements regarding these tools is a key step when developing guidelines for their use.  

Engage in meaningful conversations and ask probing questions to uncover how your team is currently using these models. This insight will allow you to tailor best practices that maximize their utility while ensuring they remain relevant and compliant within your organization.

To get this information, you may roll out a survey to understand how your team’s been using ChatGPT and what they’ve learned from it. Make sure you ask probing questions that help you understand the challenges they’ve encountered while using this technology. 

If you believe the team might be hesitant to share this information, emphasize that learning from these challenges is a natural part of the process and often leads to substantial benefits. 

3. Leverage available resources: Before huddling with your team, familiarize yourself with resources that can aid in establishing robust ChatGPT use guidelines.

OpenAI provides a valuable “Best Practices” page, which we recommend exploring. Within this resource, you’ll find a dedicated section on “safety best practices,” which delves into critical topics such as employing human verification for model outputs and addressing potential limitations like the generation of biased, offensive, or inaccurate content. By leveraging these resources, you can equip yourself with up-to-date information and guidance to share with your team.

Most executives belong to executive networks where they connect via WhatsApp or other channels. These are also good places to look for references or anecdotes. 

4. Establish practices that will foster a culture of responsible use: Focus on nurturing a workplace atmosphere where all team members are encouraged to engage with AI in a responsible and ethical manner. Considerations to take into account include:

• Promoting a verification mindset: It’s crucial to acknowledge that ChatGPT is not infallible. Approach it with the same thoroughness as a journalist would with a source, verifying information to ensure accuracy.
• Safeguarding confidentiality: Uphold the principle of confidentiality when using ChatGPT. Refrain from sharing sensitive or confidential information, as using the model is akin to making it accessible on the internet for public viewing.
• Exercising discernment with online information: Remind team members that not all information found online automatically qualifies as good, ethical, legal, or beneficial to the company. Exercise discernment when assessing the relevance and reliability of online content before incorporating it into your business operations.
• Establishing a culture of honesty about using ChatGPT or any other chatbots so it’s easier for the wider team to assess and verify this information if needed. Do this by asking the team to add the source of their data using “source: ChatGPT”.
• Defining restricted scenarios: Clearly outline situations where the use of chatbots is prohibited, ensuring that employees are well-informed about when it is strictly forbidden to use them.

Building your deck on ChatGPT guidelines: 

1. Start with a compelling example: Kick off your presentation with a relevant example to introduce the topic effectively. Choose an example that strikes a balance between highlighting the potential benefits of language models like ChatGPT for your team and acknowledging any potential challenges or risks. This approach fosters an environment of awareness without causing undue concern. Consider referencing successes like Salesforce’s Einstein or cautionary tales featured in this article to make the concept relatable.
2. Incorporate the 5W approach: Ensure a comprehensive understanding by including the 5W approach – who, what, when, where, why & how – when discussing ChatGPT’s optimal use in your company. This framework will guide your presentation, answering crucial questions about the purpose, content, timing, stakeholders, implementation, and method of using ChatGPT effectively within your organization.
3. Develop real-world scenarios: Complement your guidelines with real-world scenarios that team members may encounter in their daily work. By presenting practical examples and case studies, you can help your team visualize how ChatGPT guidelines can be applied in specific situations. These scenarios can further reinforce the importance of following the guidelines and showcase the tangible benefits of responsible ChatGPT usage. It’s also a learning tool that tends to yield better results.
4. Highlight ethical considerations: Dedicate a section of your presentation to ethical considerations when using ChatGPT. Emphasize the importance of avoiding biased, offensive, or harmful outputs and discuss strategies to mitigate these risks. Encourage team members to approach ChatGPT with a strong ethical mindset, reinforcing the responsibility of using AI in a way that aligns with your organization’s values.
5. Provide access to resources: Ensure that your team has easy access to resources for ongoing learning and support. Point them to relevant documentation, training materials, and external sources that can deepen their understanding of ChatGPT’s effective use. A well-informed team is more likely to implement these practices effectively.
6. Set clear expectations: Conclude your presentation by summarizing the key takeaways and setting clear expectations for your team. Outline what is expected of each team member in terms of ChatGPT usage and compliance with the established guidelines. Clarity in expectations helps everyone align their efforts and responsibilities.
7. Establish regular updates: Communicate your intention to revisit and update the ChatGPT guidelines periodically. Technology evolves, and so do the best practices surrounding it. Let your team know that these guidelines are not static; they’ll adapt to changes and emerging insights in the field of AI. Ask your team to have a self-educating approach. 
8. Open Q&A session: Dedicate some time to an open question-and-answer session where team members can seek clarification on any aspect of ChatGPT’s use. Encourage active participation and ensure that all concerns are addressed, further fostering a culture of transparency and understanding.

If the meeting seems to be taking longer than anticipated, you may break it down into shorter sessions. Provide team-wide visibility by sharing the deck with all concerned parties so they can refer back to it whenever needed. 

We’ve also written an article on ChatGPT outperforming human jobs, executive assistants specifically, that you may want to read.